While high-end evaluation methods like randomizedcontroltrials (RCTs) are the surest means of determining what interventions in fact cause certain outcomes, they are also time-consuming and costly and beyond the means of many, if not most, mainstream nonprofits.
These are the gold standard randomizedcontroltrials or intervention studies, in which a control group consuming soda is monitored for weight gain over time against a group which stops drinking soda (and all other dietary and lifestyle factors are controlled for).